Facebook and the Effectiveness of Election Engineering

Since November 8, 2016, we regularly hear the narrative about the Russians hijacking the presidential election through the use of fake news on social media platforms, a process that I am terming "election engineering".  Let's look at some background on election engineering from Facebook first followed by a very interesting case study showing how an election was successfully engineered using social media.

A 2017 study by Hunt Allcott and Matthew Gentzkow entitled "Social Media and Fake News in the 2016 Election" observed that social media users can share news content widely with no significant fact-checking, editorial judgement or third-party filtering.  This is particularly interesting given the following:

1.) 62 percent of U.S. adults get their news from social media

2.) the most popular fake news stories were more widely shared on Facebook than the most popular mainstream media news stories

3.) many people who see fake news satires report that they believe these stories

4.) the most discussed fake news stories tended to favour Donald Trump over Hillary Clinton

A list of fake new websites on which just over half of articles appear to be false received 159 million visits during the month of the 2016 election.

To put fake news into perspective, here is a graphic showing the percentage of Americans that believe a selection of historical partisan conspiracy theories:

Either we are a very gullible bunch of people or there are a lot of things (i.e. overwhelming government secrecy and a general distrust in politicians) that cause us to disbelieve Washington's narrative on a wide range of issues.

According to the New York Times, the bastion of media truthiness, during the 2016 election, Russians turned most often to Facebook to stir up discontent among American voters.  Here is a posting from the New York Times Facebook page regarding Russian financing of an ad linking Hillary Clinton to Satan:

Here's the ad in question:

In September 2017, Facebook disclosed that it had identified more than 3000 advertisements that had been purchased by a Russian company with alleged links to the Kremlin as shown here:

Approximately 470 inauthentic accounts were used with total ad expenditures of approximately $100,000.  In a deeper review which looked at advertising bought from accounts with American IP addresses but with the language set to Russian; in that search, they found roughly $50,000 in potentially politically-related ad spending on approximately 2200 ads.

In fact, Facebook's hunt for Russians continues as shown on this news item from July 31, 2018:

Here is a posting from one of the new fake accounts:

According to Facebook, the parties responsible for the recent fake accounts that are attempting to mold the minds of Americans have become much smarter than the Russian-based Internet Research Agency that was allegedly responsible for at least some of the 2016 election meddling violations.

Now, let's look at the surprising facts behind election engineering and social media, particularly Facebook.  Back in August 2011, a case study entitled "Reaching Voters with Facebook Ads" was released for public consumption.  In this case study, the authors looked at a ballot proposition which was held in Florida in November 2010.  The proposition itself would have seen class sizes in Florida increase.  A campaign sponsored by a group called "Vote NO on 8" focussed on defeating to proposition but had a very small budget and needed to maximize the effectiveness of its marketing to persuade voters to vote no on proposition 8.  Here's a quote from the article showing how the group got their message across to voters prior to the November 2010 election:

"The first goal of the Facebook Ads campaign was to use Facebook as a market research tool to hone the messages identified by a baseline poll specifically for each micro-audience of targeted voters in Florida and for each demographic group. The learnings from this market research would be used across all other media buys. The second goal was to saturate Facebook users in Florida with targeted messages in the month prior to the election. The third, and most important goal, was to measure the impact of the online ad program to assess its viability as a new model for voter persuasion.

Successful political campaigns often depend on repeating their message to voters often and convincingly during the period shortly before the election. Mindful of this trend, the “Vote NO on 8 in Florida” campaign ran for two months starting in September 2010. The month prior to the November election was when the campaign focused the majority of its impressions. Chong & Koster, a media firm specializing in digital communications, used Facebook’s targeting capabilities to serve a variety of ad messages to Floridians by age groups – 18 – 29; 30 – 44; 45 – 54; 55 – 63; 64 and over – separately for both males and females.

The agency relied on Facebook’s Location Targeting to reach people in two of the most populated counties in Florida, Dade and Broward, which have a combined population of 4.2 million. It chose to focus the impressions here because it wanted to be able to benchmark voter results against the rest of the state. “The methodology for using Facebook as a market research tool is really quite simple and incredibly efficient,” says Tyler Davis, Partner at Chong + Koster. “For each target audience identified by the poll, we ran a set of Facebook Ads that split-tested a variety of messages and imagery.

The agency also used Facebook to target people who liked politically oriented Facebook Pages or who listed relevant Likes & Interests or Education & Work. For example, it targeted people who listed terms like “teacher,” “pta” “math teacher” to reach educators. Because both the polling and Facebook research indicated that the issue carries special resonance with parents of school children, it even included interests like “I love my son” and “I love my daughter” (and layered them with demographic targeting.)…

Chong & Koster used Facebook’s real-time feedback capability to test multiple marketing messages with each target group. More importantly, it was able to test which images pair best with each message and audiences, which was essential in scripting the display ads. The agency broke down the response rates for each message and demographic group, and picked the most effective messages for each."

Here are the results of the Facebook political campaign:

1.) There was a 19 percent difference in the way people voted in areas where Facebook Ads ran versus areas where the ads did not run

2.) In the areas where the ads ran, people with the most online ad exposure were 17 percent more likely to vote against the proposition than those with the least

3.) Ad exposure had a greater impact on how voters voted than ideology: audiences exposed to the Facebook Ads outperformed even Democratic base voters

4.) Among people who voted against the proposition, there was a very high recall rate of the messages in the Facebook Ads, reaching as high as 45 percent or more

5.) The ads saw 75 million impressions among people in key geographic areas of Florida, likely resulting in the average Facebook user in the key areas of Florida seeing a targeted ad five times per day

Here's the conclusion:

"Chong & Koster found that achieving such dramatic results was very economical using Facebook. “The beautiful value in this is that we can actually move public opinion intentionally and cheaply. The amount spent on Facebook buys is the amount it would cost to put one mailer in 150,000 voters’ mailboxes,” says Josh Koster, partner at Chong and Koster. “If one-third of the voters read the mailer, that’s equivalent to 50,000 voters for the same amount of money that we used to reach to the two largest counties in Florida at frequency for months.” (my bold)

Isn't it interesting to see that, way back in 2010, Facebook was being used to sway public opinion during an election (in other words, election engineering).  You might ask, where did this study appear?  Here's a link to the case study and here are two screen captures showing the entire case study webpage (just in case Big Brother takes it down):

That's right, the Facebook election influencing case study appears on Facebook's own Government and Politics on Facebook page in a posting dated August 16, 2011.

This begs the question, did Facebook's own case study for election meddling form the basis for Russia's alleged election meddling in 2016?  In both cases, the possibility that social media platforms can be used to engineer elections says a lot about the current state of news literacy in the United States and its not particularly good news.